“Making the IMAGE model FAIR is a major step for science and policy”

In this series of interviews we show what contribution projects can make to FAIR research IT. The research teams of the projects have received a grant from the FAIR Research IT Innovation Fund.

The IMAGE model is used to research global warming and its countermeasures. This computer model is not open-source and the computer code is outdated. Thanks to the Innovation Fund, the part on land use, IMAGE land, has been converted into an open-source Python-version. This is one of the first steps in a larger transition to make the IMAGE model more FAIR, publicly available and transparent. Utrecht University researchers collaborated on this project with fellow applicant Netherlands Environmental Assessment Agency (Planbureau voor de Leefomgeving) (PBL).

Climate change and land use generate plenty of public and scientific interest. As a scientist, Dr. Judith Verstegen explores how best to explore these topics, both methodologically and content-wise. Judith is an assistant professor in geo-information science at the department of Human Geography and Spatial Planning. “My expertise is geographic information science, GIS. I develop methods for geographical research. My focus is on simulation models and spatial optimisation, where I am looking at spatial systems having a human component. For instance, I have recently worked on modelling pedestrian movement in cities. And during my PhD I was engaged in land use. That’s how I came to know Jonathan.” Jonathan Doelman is researcher at the Netherlands Environmental Assessment Agency (PBL) and guest researcher at the Utrecht University Copernicus Institute of Sustainable Development. “At PBL we analyse long-term scenarios in the field of climate, environment and sustainable developments: what does the future look like, what is the effect of policy measures on climate change for example? I myself look into the role of land use, such as agriculture and deforestation.”

Integrated Assessment Modelling

Doelman and his colleagues perform these scenario analyses with Integrated Assessment Models (IAMs), he explains. “With the help of this type of model we simulate interactions between the human system, such as energy use and land use, and the biophysical system, such as climate and forests.”  Verstegen adds: “ ‘Integrated’, in this case, means that all kinds of different model components are interacting. For instance, one component simulates the land system, another one climate.”  She says that an IAM does not work data driven, but theory driven. “You start with theories, like economic theories about market mechanisms. You implement those assumptions as rules in code. Next you can apply these rules to a certain case, such as a country and a time period. To apply the model you also need data, for instance from the HYDE database. Next you can run your analysis. Maybe you want to study the demand for food. Then the model checks how much land use is needed in the future and where new arable farming could be located.” Besides future scenarios, these models allow outlining what-if scenarios, by mapping the effects of possible policy interventions.

The IMAGE-model

There are several IAMs, including the IMAGE model (Integrated Model to Assess the Global Environment). The owner of this model, PBL, works with it, as well as a UU research group and partners in Wageningen and Germany. Doelman:  “We use it to analyse the big picture on climate, where are we heading in the long run? It is very useful for science on a global scale.”  The IMAGE model therefore plays a major role in large analyses such as those of the Intergovernmental Panel on Climate Change (IPCC) and The Intergovernmental Science-Policy Platform on Biodiversity and Ecosystem Services (IPBES). “We analyse for instance: what should you do from a policy perspective to achieve climate goals? In doing so, I myself am looking at land use: what needs to be changed in the agricultural systems, how much bioenergy and reforestation is needed, what are the interactions with food supply, for example? I try to find answers with IMAGE land, the part of the IMAGE model that simulates land use.”

Closed source and not FAIR

In short, the IMAGE model enables interesting research that is also relevant to policy and society. Yet it has drawbacks, says Doelman.  “It is written in several programming languages, outdated and closed source, making the model inaccessible and our work less transparent.”  He says that researchers, such as PhD candidates from (foreign) institutes, regularly ask where they can download and use the IMAGE model.  “ We have to disappoint them. Moreover, being closed source makes it harder for IMAGE users to get their research published. Such a shame.”

That is why the IMAGE team is in the process of making the model FAIR, says Doelman. “We are in a larger transition to open source and better documentation. We also want to convert the code to one common programming language: Python. That will make the model much more accessible and future-proof.”  He explains that some other IAMs are already open source by now. “As a result, the number of researchers working on it increases rapidly. That is the interesting thing about FAIR and open source science: you can work with more people on a research topic, in a structured way. You also continue to develop faster, because others add things to your model. Verstegen: “Plus: if more people use your model, you discover bugs sooner. Important, especially if it is used to develop policy. Open source greatly increases reproducibility.”  Finally it provides more transparency. Doelman:  “It is important that people reading about IMAGE or reviewing our work can have a look at the code. Then they understand what is happening. And if they are asking critical questions? Then we can improve the model, leading to even better research.”

That is the interesting thing about FAIR and open source science: you can work with more people on a research topic, in a structured way.

Consortium

In this larger transition Doelman was looking for ways to make ‘his’ part, IMAGE land, FAIR. “But because that is such an enormous job I needed help. That is how I got in touch with Judith.” Verstegen:  “Jonathan approached me already before the Innovation Fund call. I really loved to help, but it was too time-consuming. When I saw the call I thought: what a great opportunity to submit an application together. As a university it is also good to support this kind of initiatives coming from research institutes.”  In addition, she had already developed an open source land use model during her PhD. “ So I could add some of my knowledge from that experience.”

Verstegen became the lead applicant and brought together a consortium. “We collaborated with three UU departments: Human Geography, the Copernicus Institute and Physical Geography. From PBL a few other researchers were involved besides Jonathan Doelman.” The budget allowed them to hire a researcher at the Human Geography department (Ben Romero-Wilcock)  for a seven-month-period to convert IMAGE land to an open-source Python code.  Verstegen did the daily supervision, Doelman answered questions about the IMAGE model. “The budget also allowed us to bring in an engineer from Physical Geography for technical support and to organise two workshops. “In the kick-off workshop we brought everybody together and formulated criteria for the ‘new’ model. In the final workshop we demonstrated the software to PBL researchers.” Verstegen says that the collaboration with PBL was ideal for developing software that is best suited to practice. “Of course you want an improved model to meet the needs and wishes of those who are going to work with it. As PBL was one of the end users, this collaboration proved very useful.”

Thanks to this collaboration we could develop software that is best suited to practice.

Interesting research

Besides PBL researchers, UU researchers also benefitted from making the IMAGE model FAIR. Doelman: “The collaboration was already close, UU professor Detlef van Vuuren is also senior researcher at PBL and leads the IMAGE team that includes both PBL researchers and UU researchers.” In addition, the project is of value to other (future) IMAGE users. Doelman: “The new version is of course useful for new members of the IMAGE team, but maybe more so for people from outside the team. As a result, all parties interested can better understand and use the model.  Researchers working with other models also benefit from having made IMAGE FAIR, says Verstegen.  “For instance, people are increasingly comparing IAMS or running them together to calculate uncertainty ranges. Others can now do these kinds of studies much easier with the IMAGE model.”

Wider impact

Besides science, also policy and society profit from the FAIR IMAGE model. Doelman:  “Think for instance of the political discussions about models that are used by planning agencies and knowledge organisations, such as the nitrogen model on which much of the nitrogen policy is based. The more influence models have on policy – and so on people’s lives -  the more important it is that you are able to explain clearly what the model does. Society benefits from quantitative calculations, but in doing so we have to be transparent.” Verstegen also thinks that projects like these help to increase understanding and trust among citizens.  “A transparent model makes it easier to accept the consequences. For example, the first model that calculated the scenarios during the pandemic was closed source. It received a lot of criticism, it damages trust. If the software is open, not everybody will take a look at everything, but media do write about it and in their turn influence people. It is often thought, and rightly so: if it is not transparent, can it be trusted? This is another reason why working FAIR is of major importance.”

Significance of the project

Doelman says that the Innovation Fund was an important step for the IMAGE team in the larger transition to open-source and Python. “It was a very useful process and a good collaboration with the UU. The part that is finished is already in Github. Unfortunately, the IMAGE model as a whole cannot be used yet, due to its size not everything has been converted yet.” Verstegen explains:  “We made sure that the land use component can also be run separately, so for land use the model is already accessible and FAIR. “ For her, the project fitted in nicely with her own research.  “I practise science to help people and to move developments forward. I am also proud of the broad collaboration. At UU, geo information scientists do not share the same workspace, but are spread across different departments within Geosciences.That is why people sometimes forget that also methodological research on spatial analysis methods or software development is being carried out in faculty too. That is what I am trying to make more visible. This project has contributed to that goal.”